YOLO-NAS

YOLO-NAS is an object detection model developed by Deci that achieves SOTA performances compared to YOLOv5, v7, and v8.

Overview

YOLO-NAS is an object detection model developed by Deci that achieves SOTA performances compared to YOLOv5, v7, and v8.

Performance

Model	mAP	Latency (ms)
YOLO-NAS S	47.5	3.21
YOLO-NAS M	51.55	5.85
YOLO-NAS L	52.22	7.87
YOLO-NAS S INT-8	47.03	2.36
YOLO-NAS M INT-8	51.0	3.78
YOLO-NAS L INT-8	52.1	4.78

mAP numbers in table reported for Coco 2017 Val dataset and latency benchmarked for 640x640 images on Nvidia T4 GPU.

Use This Model

First, install Inference:‍


pip install inference

To try a demo with a model trained on the Microsoft COCO dataset, use:


import inference
model = inference.load_roboflow_model("yolo-nas-s-640")
results = model.infer(image="YOUR_IMAGE.jpg")

Above, replace:

YOUR_IMAGE.jpg with the path to your image.

You can also run fine-tuned models with Inference.

Retrieve your Roboflow API key and save it in an environment variable called ROBOFLOW_API_KEY:


export ROBOFLOW_API_KEY="your-api-key"

To use your model, run the following code:


import inference
model = inference.load_roboflow_model("model-name/version")
results = model.infer(image="YOUR_IMAGE.jpg")

Above, replace:

YOUR_IMAGE.jpg with the path to your image.
model_id/version with the YOLO-NAS model ID and version you want to use. Learn how to retrieve your model and version ID.

Label Data Automatically with YOLO-NAS

You can automatically label a dataset using YOLO-NAS with help from Autodistill, an open source package for training computer vision models. You can label a folder of images automatically with only a few lines of code. Below, see our tutorials that demonstrate how to use YOLO-NAS to train a computer vision model.

OWL-ViT and YOLO-NAS

Grounding DINO and YOLO-NAS

SAM-CLIP and YOLO-NAS

Grounded SAM and YOLO-NAS

DETIC and YOLO-NAS